README.md

Overview

Design Deck is an open-source collection of 230+ system design flash cards.

It helps you prepare and succeed in your system design interview.

The topics covered are the following:

• Cache: eviction, locations, strategies, when to use a cache, etc.
• Cloud: main cloud components
• Database: ACID, CAP, partitioning, consistency, isolation, denormalization, etc.
• Design: general topics such as idempotence, bloom filter, causality, asynchronous communications, vector clocks
• HTTP: main HTTP knowledge
• Kafka: main Kafka building blocks
• Math: discrete math
• Network: TCP, CORS, ping & heartbeat, OSI, etc.
• Reliability: reliability patterns
• Security: main security knowledge such as TLS, cipher, encryption

Anki Deck

Anki is a free software (Windows/Mac/Linux/iPhone/Android) which makes remembering things easy. It utilizes spaced repetition which is a proven technique to increase the rate of memorization:

Spaced Repetition: The most powerful study technique on YouTube

The single biggest change that Anki brings about is that it means memory is no longer a haphazard event, to be left to chance. Rather, it guarantees I will remember something, with minimal effort. That is, Anki makes memory a choice.

Michael A. Nielsen, "Augmenting Long-term Memory"

Using Anki is a great way to prepare your system design interview. Here is a flashcard example:

The Anki version (a clone of the +200 flashcards from this repo) is available via a one-time GitHub sponsorship tier for $21: ❤️ Sponsor, One-time tab, Access to the latest Anki deck version of Design Deck tier.

Cards Index

Cache

• Cache aside
• Cache aside vs. read-through
• Cache eviction policy
• Cache locations
• Cache: refresh-ahead
• Cache: write through vs. write back
• Four main distributed cache benefits
• Main metric for cache
• Read-through cache
• When to use a cache

Cloud

• CDN

DB

• 3 main reasons to partition data
• ACID property
• Anti-entropy
• Byzantine fault-tolerant
• CALM theorem
• CAP theorem
• Caveat of serializability
• Chain replication
• Chain replication vs. consensus
• Change data capture (CDC)
• Concurrency control
• Consensus
• Consistency models
• CQRS
• CRDT
• CRDT and collaborative applications (e.g., Google Docs)
• DB indexes tradeoff
• DB internal components
• DB: read vs. write-heavy, latency vs. consistency, availability vs. consistency, ACID vs. non-ACID
• Delta CRDTs
• Denormalization
• Design consideration when partitioning data
• Downside of distributed transactions
• Event sourcing
• Eventual consistency requirements
• Examples of solutions offering leader election abstractions
• Federation
• Fencing token
• Gossip protocol
• Graph DB main use case
• Hinted handoff
• Hot spot in partitioning
• In a database, strategy to handle rebalancing
• Isolation levels
• Known CRDTs
• Last-write-wins (LWW)
• Leader election
• LSM tree
• LSM tree vs. B-tree
• Main difference between consistency models and isolation levels
• Merkle tree
• Monotonic reads consistency implementation
• MVCC
• N+1 select problem
• NoSQL: main types and main architecture principles
• Operation-based CRDTs
• Operational transformation (OT): concept and main drawback
• Optimistic concurrency control: pros and cons
• PACELC theorem
• Partitioning (sharding)
• Partitioning criteria
• Partitioning methods
• Quorum
• Raft
• Read repair
• Relation between replication factor, write consistency and read consistency
• Replication vs. partition: impacts
• Schema-on-read vs. schema-on-write
• Serializability
• Serializable Snapshot Isolation (SSI)
• Single-leader, multi-leader, leaderless replication
• Sloppy quorum
• Snapshot Isolation (SI)
• Snapshot Isolation common implementation
• SSTable
• State-based CRDTs: definition and requirements
• Strong eventual consistency: definition and requirements
• Three-phase commit (3PC)
• Transaction
• Two main approaches to partition a table that has secondary indexes
• Two types of CRDTs
• Two-phase commit (2PC)
• WAL
• When relational vs. when document
• When to use a column-oriented store
• Why DB schemaless is misleading
• Why is in-memory faster
• Write and read amplification
• Write heavy and replication type

Design

• Auditing
• Backward vs. forward compatibility
• Bloom filter
• Causality
• Concurrent operations
• Consistent hashing
• Design impacts of sharing
• Design: read-heavy vs. write-heavy impacts
• Different types of message failure
• Event log vs. message queue
• Exactly-once delivery
• FLP impossibility
• Geohashing
• Hashing definition and size of MD5 and SHA256
• HDFS
• How to reduce sharing
• HyperLogLog
• Idempotent
• Latency numbers every programmer should know
• Lease
• Least loaded endpoint load balancing strategy
• Liveness property
• Load balancing
• Load shedding
• Locality
• Log
• Log compaction
• Main drawback of shared-nothing architectures
• MapReduce
• Microservices: pros and cons
• Number of values to generate to reach 50% chances of collision: 32-bit, 64-bit, and 128-bit hash
• Orchestration vs. choreography
• Outbox pattern
• Perfect hashing
• Quadtree
• Rate-limiting (throttling): definition and algos
• Rebalancing
• REST
• REST vs. gRPC
• Safety property
• Saga
• Scalability
• Scalability ceiling
• Shared-nothing architectures
• Source of truth
• Split-brain
• Throughput
• Total vs. partial order
• UUID
• Validation vs. verification
• Vector clock
• Why asynchronous communication

HTTP

• 301 vs. 302
• 403 or 404?
• Cookie
• Four main HTTP/2 features
• HLS
• HTTP
• HTTP cache-control header
• HTTP Etag
• HTTP keep-alive
• HTTP methods: safeness and idempotence
• HTTP safe method
• HTTP status code 429
• HTTP status codes
• What happens if you type google.com in your browser

Kafka

• Consumer types
• Durability/availability and latency/throughput tradeoffs
• Log compaction
• Offset
• Partition
• Partition distribution
• Rebalancing
• Segment
• Shared subscription

Math

• Associative property
• Commutative property
• Harmonic mean

Network

• ARP protocol
• Average connection speed in USA
• Backpressure
• Bandwidth
• BGP
• CORS
• Difference ping & heartbeat
• Difference TCP & UDP
• Difference view & materialized view
• DNS
• DNS lookup: push or pull
• Health checks: passive vs. active
• Internet model
• Layer 4 vs. layer 7 load balancer
• MAC address
• Max size of a TCP packet
• MQTT LWT
• NTP
• OSI model
• Routers
• Routers buffering
• Routers processing
• Routing table
• Service mesh
• Switch
• TCP congestion control
• TCP connection backlog
• TCP flow control
• TCP handshake
• Websocket
• Why can't we rely on the system clock in distributed systems?

Reliability

• Bulkhead pattern
• Cascading failure
• Causal consistency implementation
• Circuit breaker
• Exponential backoff
• Fault tolerance
• Jitter
• Knee point
• Phi-accrual failure detector
• Retry amplification

Security

• Authentication
• Certificate authorities
• Cipher
• Confidentiality
• Integrity
• Mutual TLS
• OAuth 2
• Public key infrastructure (PKI)
• TLS handshake
• Two main uses of encryption
• Two types of encryption
• What does digital signature provide
• What does TLS provide?

References

• Designing Data-Intensive Applications
• Understanding Distributed Systems
• Distributed Systems for practitioners

Additional Notes

If you're preparing an algorithm & data structure interview, you can take a look at Algo Deck